Plant Syst. Evol. 225:43-72 (2000) Plant Systematics and Evolution © Springer-Verlag 2000 Printed in Austria Phylogeny and classification of the subfamily Rubioideae (Rubiaceae) B. Bremer 1 and J.-F. Manen 2 1Department of Systematic Botany, EBC, Uppsala University, Uppsala, Sweden 2Universit6 de Gen6ve, Conservatoire et Jardin Botaniques, Chamb&y, Switzerland Received April 27, 1999 Accepted June 21, 2000 Abstract. We performed phylogenetic analyses of Progress in understanding of the subfamily the subfamily Rubioideae (Rubiaceae) based on Rubioideae of the Rubiaceae is relatively three different pieces of chloroplast DNA, the recent and includes many important contribu- protein coding rbcL gene, the spacer sequence tions from many different scientists. Before the between atpB and rbcL (atpB-rbcL), and the recently middle of the 20th century the "Rubioideae" published (Andersson and Rova 1999) rpsl6 intron taxa were dispersed in the two subfamilies data. New rbcL sequences have been produced for Coffeoideae and Cinchonoideae, a classifica- 41 taxa and there are 52 new atpB-rbcL spacer sequences. All analyses gave similar results concern- tion of the Rubiaceae based on ovule number ing the phylogeny, but they differ slightly in reso- (Schumann 1891). Bremekamp (1952, 1954) lution and support for the various branches. The and Verdcourt (1958) argued against this minor tribes Ophiorrhizeae, Urophylleae, Lasian- artificial division of the family and instead theae, and Coussareeae form a grade to the rest of proposed that all Rubiaceae tribes with species the subfamily, which consists of two well-supported containing raphides (calcium oxalate crystals) branches, the Psychotrieae alliance and the Sper- should be set aside as a new subfamily, macoceae alliance, including a majority of all genera Rubioideae. There are large similarities and species. Based on the resulting phylogenies we between the systems of these two authors. present a revised classification of the Rubioideae. Verdcourt included 16 tribes (and the unplaced We accept 16 tribes of which 12 more or less genus Hillia) in the subfamily Rubioideae, and correspond to earlier tribal circumscriptions: Anthospermeae, Argostemmateae, Craterisper- Bremekamp (1966), who used a more narrow meae, Gaertnereae, Morindeae, Paederieae, tribal circumscription, included 19 tribes and Psychotrieae, Schradereae, Spermacoceae, Rubieae, excluded some tribes to other subfamilies: Theligoneae, and Urophylleae; two tribes have Hillieae to Hillioideae, Ophiorrhizeae to received new and very different circumscriptions: Ophiorrhizoideae, and Urophylleae and Ophiorrhizeae and Coussareeae; and two are new to Pauridiantheae to Urophylloideae. Characters, science: Lasiantheae and Danaideae. except the raphides, that were used to distin- guish Rubioideae were valvate aestivation Key words: Rubiaceae, Rubioideae, Danaideae, (with few exceptions, Harnelia, Deppea, and Lasiantheae, chloroplast DNA, rbcL, atpB-rbcL Hillia - all these have later been shown to spacer, rpsl6 intron, phylogeny, classification. belong to Cinchonoideae; e.g. Bremer et al. 44 B. Bremer and J.-F. Manen: Rubioideae classification 1995) and indumentum of septated (articulat- (Bremer and Jansen 1991) and Hillieae (Bremer ed) hairs. The next major change in the et al. 1995, Natali et al. 1995) do not belong in Rubioideae circumscription was presented by the subfamily, that Theligoneae belong in the Robbrecht (1988). His Rubioideae with 16 Rubioideae close to Rubieae (Bremer 1996a, tribes agreed largely with Bremekamp's but Natali et al. 1996), and that Knoxieae belong Robbrecht included Ophiorrhizeae and Theli- in the Rubioideae and are not part of Anti- goneae (shown to be related to Rubiaceae by rheoideae (Bremer 1996a). Wunderlich 1971) and excluded Knoxieae Molecular investigation has proved to be and Craterispermeae to the subfamily very informative for reconstruction of phylo- Antirheoideae. These three classification geny but several tribes are yet to be represented schemes of Bremekamp, Verdcourt, and Rob- in molecular investigations. Just before brecht have been important sources of inspi- completion of this manuscript a detailed ration for later more focused studies of analysis of the Rubioideae based on the rpsl6 taxonomy and morphology. Comprehensive intron was published (Andersson and Rova taxonomic studies of different Rubioideae 1999). That study confirms, in most parts, tribes that should be mentioned are those by earlier molecular studies (Bremer 1996a, Natali Puff and Robbrecht with co-authors (Puff et al. 1996) but as more taxa are included it 1982, 1988; Puff and Robbrecht 1989; Puff also contributes a lot of new information. and Buchner 1998; Robbrecht 1982; Robb- The present study aims at presenting new recht et al. 1991), and Johansson (1987a, b, phylogenetic analyses of the Rubioideae based 1988, 1989, 1994). Other examples are chro- on two different sequences of chloroplast mosome investigations by Kiehn (e.g. 1986, DNA, the protein coding rbcL gene and the 1995), morphological studies by Igersheim and spacer sequence between atpB and rbcL co-authors (Igersheim 1992, Igersheim and (atpB-rbcL), a combined analysis of these Rohrhofer 1993, Igersheim et al. 1994), and data, and also an analysis of our data in pollen investigations by e.g. Robbrecht and co- combination with the recently published rpsl6 workers (Robbrecht 1982, Jansen et al. 1996). data, rbcL/atpB-rbcL/rpsl6 (also referred to as The classification systems as well as the the 3-data matrix or analysis). As a result of taxonomic/morphological investigations have the phylogenetic analyses we present a new in their turn inspired phylogenetic investiga- classification of the subfamily Rubioideae. tions based on molecular data. In many of these, Rubioideae taxa have been included even if the focus has been on other groups or at Material and methods other taxonomic levels. Several studies have The sampling strategy in the phylogenetic analyses aimed at resolving the whole family phylogeny was to cover as many of the described tribes of the (Bremer and Jansen 1991, Bremer et al. 1995, subfamily Rubioideae as possible, and also to Bremer 1996a), others at specific tribes, e.g. include several taxa from large and diverse tribes. Rubieae (Manen etal. 1994, Manen and We did not manage to find material from two of the Natali 1995, Manen and Natali 1996, Natali twenty-eight tribes (Lathraeocarpeae and Pera- et al. 1996), and Psychotrieae (Nepokroeff meae) and for economical and technical reasons et al. 1999). These studies are based on mo- we did not sequence all taxa for both rbcL and the spacer. All 151 genera included in the resulting lecular markers designed to elucidate phylog- classification are listed in the appendices 1 and 2. enies at different taxonomic levels but the The outgroups were chosen to represent one results are congruent to a great extent. The Gentianales member outside the Rubiaceae, results in some parts agree with classification Gelsemium (Gelsemiaceae; in the spacer and com- but in other parts they are totally different and bined analyses), and members of other subfamilies there are several important conclusions of the Rubiaceae (cf. Bremer 1996a), Gardenia from concerning Rubioideae, e.g. that Hamelieae the subfamily Ixoroideae, and Cinchona and B. Bremer and J.-F. Manen: Rubioideae classification 45 Luculia (in the spacer analysis also Coptosapelta were without indels and the alignments were simple and Exostema) from the subfamily Cinchonoideae. and without any alternatives. In the analysis of our data in combination with the The parsimony analyses were done using rpsl6 data we used Gelsemium and Cinchona as PAUP 3.1 (Swofford 1993). Only phylogenetically outgroups. informative characters were included. The search Silica gel-dried or herbarium leaves were used method for the combined matrices was heuristic in the DNA investigations. DNA was extracted, with 100 replications of RANDOM stepwise addi- amplified, and sequenced according to Bremer et al. tions of sequences, the TBR branch swapping, and (1995) and Backlund et al. (2000) for rbcL, and MULPARS options in effect. Support of the clades according to Manen et al. (1994) for the atpB-rbcL was calculated with bootstrap analyses with 1000 spacer. Table 1 provides a list of taxa included in replicates and TBR branch swapping and MUL- the phylogenetic analyses of rbcL and atpB-rbcL PARS off. data, with EMBL/GenBank accession number of the corresponding sequences, and with voucher information for new sequenced taxa. New rbcL Results sequences have been produced for 41 taxa and there The atpB-rbcL spacer matrix (63 taxa) com- are 52 new spacer sequences. All other atpB-rbcL prises 970 aligned positions, 453 of which are sequences used in these analyses have been pub- variable and 255 are potentially informative, 35 lished earlier by the first or second author alone or of which are indel characters. The matrix in collaboration with co-authors (Olmstead et al. rbcL 1993; Bremer et al. 1995; Bremer 1996a, b; Andrea- (106 taxa) comprises 1 402 sites, 560 of which sen and Bremer 1996; Manen et al. 1994; Manen are variable and 404 are potentially informa- and Natali 1995, 1996). For references to the rpsl6 tive. The combined rbcL/atpB-rbcL matrix (59 data see Andersson and Rova (1999). taxa) comprises 2 364 sites, 899 of which are The rbcL and atpB-rbcL spacer matrices com- variable and 533 are potentially informative. prise 106 and 63 taxa, respectively. The combined The combined rbcL/atpB-rbcL/rpsl